ICAP 4 th Workshop, 14-17 May 2012, ESRIN, Frascati, Italy The MACC/ECMWF aerosol analysis and forecast system: Recent results Jean-Jacques Morcrette ECMWF, Reading UK Angela Benedetti, aerosol analysis Johannes W. Kaiser , everything fire-related Luke Jones, AERONET, CALIPSO visualisation and some more diagnostics Miha Razinger , MACC web site and some other diagnostics Martin Suttie data acquisition and preparation

The MACC/ECMWF aerosol analysis and forecast system ...icap.atmos.und.edu/AERP/MeetingPDFs/Overviews/6.1ECMWF_Morcre… · The MACC/ECMWF aerosol analysis and forecast system: Recent

Embed Size (px)

Citation preview

ICAP 4th Workshop, 14-17 May 2012, ESRIN, Frascati, Italy

The MACC/ECMWF aerosol analysis and forecast system: Recent results

Jean-Jacques MorcretteECMWF, Reading UK

Angela Benedetti, aerosol analysisJohannes W. Kaiser, everything fire-relatedLuke Jones, AERONET, CALIPSO visualisation and some more diagnosticsMiha Razinger, MACC web site and some other diagnosticsMartin Suttie data acquisition and preparation

The MACC / ECMWF aerosol analysis and forecast system

• ECMWF for its operational analyses and forecasts ingest gigabytes of observational data every day and produces weather forecasts of temperature, humidity, wind, …

• Since 2006, with the EU FP7 GEMS, then MACC projects, ECMWF has been producing experimental analyses and forecasts of the same meteorological parameters PLUS those of reactive gases, greenhouse gases, and aerosols. Near Real Time forecasts started in Sep.’08.

• GEMS produced a 2003-2008 reanalysis of GHG, RG and AER; MACC produced another reanalysis based on a more recent meteorological model and upgraded GHG, RG and AER system for 2003-2010. MACC II has been running since Nov.’11, and aim at introducing a further improved GHG, RG and AER system into operations by the end of 2014.

• This presentation will give a quick overview of the MACC aerosol system, then will briefly address recent model developments related to:– Desert dust– Volcanic aerosols– Progress in aerosol analysis (using the fine mode information, and preliminary

work on assimilating CALIPSO backscatter)

Conclusions and perspectives

MACC – Monitoring Atmospheric

Composition and Climate


• Integrates space-based and in-situ observations of atmospheric composition with state-of-the art atmospheric modelling provided by the ECMWF IFS

• Provides monitoring and forecasting services

• Helps Europe to respond to climate change and poor air quality

Since Nov’11, MACC II follow-up project

MACC Daily Service Provision

Air quality

Global Pollution


UV index

Biomass burning


NRT since September 2008



Quick overview of the MACC/ECMWF aerosol analysis and forecasting system

12 aerosol-related prognostic variables added

to the existing prognostic variables (T, wind, q,

O3, Ps)

* 3 bins of sea-salt (0.03 – 0.5 – 0.9 – 20 µm)

* 3 bins of dust (0.03 – 0.55 – 0.9 – 20 µm)

* Black carbon (hydrophilic and –phobic)

* Organic carbon (hydrophilic and –phobic)

* SO2 -> SO4

Physical processes include:

•emission sources (some of which updated in

NRT, i.e.fires),

•horizontal and vertical advection by dynamics

•vertical advection by vertical diffusion and


•aerosol specific parameterizations for

dry deposition, sedimentation, wet deposition

by large-scale and convective precipitation

(SS, DU, OM, BC, SU) and hygroscopicity

(SS, OM, BC, SU).

Forward model

Integrated in the ECMWF incremental 4D-Var

Control variable is formulated in terms of the

total aerosol mixing ratio. Soon to come: fine

and coarse mode. Increments in total mass

are repartitioned into the single species

according to their fractional contribution to the


Background error statistics have been

computed using forecasts errors as in the

NMC method (48h-24h forecast differences).

Assimilated observations are the MODIS

Aerosol Optical Depths (AODs) at 550nm over

land and ocean. Observation errors, originally

prescribed, are now provided as part of the

variational bias correction.


Morcrette et al., 2009, JGR Benedetti et al., 2009, JGR

Evaluation with MODIS/SEVERI and AERONETSaharan dust outbreak: 6 March 2004

Model simulation Assimilation MODIS


Cape Verde DakarAERONET



Aerosol optical depth at 550nm (upper)and 670/675nm (lower)

Morcrette et al., 2009; Benedetti et al., 2009

Russian fires, August 2010

The MACC-AERosol system, assimilating MODIS tau550 allowed a good description of the impact of the Russian fires in mid-August 2010 over North-Eastern Europe

Helsinki Toravere Kuopio

Hyytiala Minsk Sevastopol

Ankara AERONETTotalSulphateDustOrganic CBlack C

More on fires from J. Kaiser on Wednesday

How good (or otherwise) is the MACC/ECMWF aerosol system?

ECMWF fbov: MACC Reanalysiscompared to AERONET aerosol sunphometer measurements over a year

2003 2010



New CY37R3Old CY36R1 (the MACC reanalysis)

old new

April ‘12


New CY37R3Old CY36R1 (the MACC reanalysis)

2/N* (f-o)/(f+o)

Africa Europe

N.Amer. S.E.Asia



Amsterdam Island Ascension Island

Crozet Island Guam

La Parguera Midway Island

Nauru Tahiti


Dust only20110301-20110531


Dry Deposition Sedimentation

Wet Deposition LSP Wet Deposition CP

Parameters governing dust emissions

Fraction Bare Soil



First LayerSoil Moisture

Near Real Time System



Solar Village

Sede Boker

CairoForth CreteLampedusaCabo da Roca


La Parguera

Ragged Point “Fennec”Bambey

May 2011

10 days of GEMS analyses 14-23 April 2007 10 days of MACC analyses 14-23 April 2007

10 day-FC with MACC AER 14-23 April 201010 day-FC with revised dust 14-23 April 2010

Revisions to the GEMS/MACC Aerosol Model

One deficiency in the original GEMS model was the large dust load, mainly constituted of fine (0.03-0.55 um particles. The amount produced was revised for MACC, with no real impact on the size distribution. A recent revision of the source formulation has markedly improved this aspect of the model.

MACC RevisedBin 1

Bin 2

Bin 3

Ilorin Izana

Lecce Uni Praia


Ilorin Solar Village

Tamanrasset, Dec’11 Tamanrasset, Apr.’12


Other revisions to the GEMS/MACC aerosol model

• Additional sets of optical properties have been computed for radiation diagnostics/direct effect using refractive indices from Bond &Bergstrom (2006) or Stier et al. (2007) for BC; Woodward et al. (2001) or Highwood et al. (2009) for DU.

• Lidar simulator at 355, 532, and 1064 nm has been developed (used in comparisons with CALIPSO, and for work on assimilation of lidar backscatter profiles)

• Following what is done in the M7 model:– Coefficients for dry deposition have been made function of the

underlying surface type (ocean, ice/snow, land)– Coefficients for wet deposition, sedimentation have been

adapted from M7

Daily MACC forecasts: Eyjafjallajökull eruption

The global MACC system at ECMWF provided daily 4-day forecasts of the plume shape based on basic assumptions for the injection height and mass.

April-May 2010

But the original GEMS aerosol system wasanalysing the increased optical depth from the volcanic plume and was affecting it as an increased amount of the dominant aerosol in this region, i.e., sea salt aerosol

Recent work on Volcanic aerosols in ECMWF/MACC• A new prognostic variable has been added to represent volcanic aerosols, and the code reorganised to allow volcanoes to be “switched

on or off”, with namelist or file definition of the emitted mass and boundaries of the plume. • Simulations of the 2010/04 Icelandic volcano have been performed (and a few others ...) • Sensitivity to optical properties given to volcanic aerosols (passive, SO4, BC, DU3)• Sensitivity to efficiency of gravitational deposition (sedimentation)• Sensitivity to vertical distribution of ejecta• Sensitivity of details of the profile of ejecta• 10-day forecast vs. cycling forecast forced every 12 hours by analysis

Continuously emitting Potentially explosive

UL:passive; UR: SO4; LL: BC; LR: DU3

Sensitivity to gravitational sedimentation:

Top is coefficient / 5

Middle is reference coefficient

Bottom is coefficient * 5

Sensitivity to vertical distribution of homogeneously distributed ejecta:

Top: upper half: 3350-6000 m

Middle: full layer: 700-6000 m

Bottom: lower half: 700-3350 m

UL: Stohl FC; UR: constant FC; LL: Stohl cycling; LR: constant cycling

Preliminary conclusions on the volcanic aerosol modelling work

• Each volcano is obviously different and the behaviour of the aerosol plume depends on “N” parameters!

• Over a few days (~ 2-3), dynamics govern the dispersion, sedimentation is important, optical properties are not.

• For a future operational MACC-type system (for which profile information is not likely to be available in NRT), it appears better to have a “volcano” system with only a few parameters that could be adjusted to give a reasonable agreement with MODIS-type observations of optical depth.

Summary• From the meteorological model point of view, the

new aerosol system benefits from a better cloud scheme (5 progn.variables) and other modifications to the physics package.

• A revision of the dust source formulation has corrected the strong bias on small dust particles seen before, and improved the steadiness of the aerosol load during the forecasts.

Future aerosol modelling work

• “Back-burner” work on direct/indirect effects of GEMS/MACC aerosols to (possibly) improve the FCs.

• Making volcanoes “go” at the turning of a LOGICAL

• At part of MACC II, finalise the introduction of UKCA_GLOMAP_MODE in the IFS.

• Comparison between IFS-GLOMAP and GEMS/MACC aerosols in analysis and forecast modes against observational datasets (AERONET, GAW, CALIOP, MISR, …)

Prospective/Perspectives• In forecast mode, the present (simple) prognostic aerosol system at ECMWF

adds 85% extra time and 200% extra memory to the operational configuration.

• An operational configuration therefore can only be run at relatively low resolution (presently T255 L60 vs. T1279 L91).

• Presently, aerosols are analyzed with assimilation of MODIS tau550, but the aerosol analysis does not affect the rest of the analysis (T,q, winds, O3).

• Having the aerosols affecting the full analysis system requires the development of the TL/AD of the aerosol model.

• In this case, is a more sophisticated (and potentially better) modal aerosol model affordable?

• Figure out and thoroughly test a viable (affordable) configuration of the ECMWF IFS allowing aerosol analysis and forecast to become operational (Present is TL255 L60) in 2014, for post-MACC II activities.

Present state of the prognostic aerosols in the ECMWF IFS

• Benedetti et al., 2009: JGR, 114, D13205• Morcrette et al., 2008: GRL, 35, L24813• Morcrette et al., 2009: JGR, 114, D06206• Mangold et al., 2011: JGR, 116, D03302• Kaiser et al., 2009: AIP CP, 1100• Kaiser et al., 2010: BAMS, 91, SC2009• Kaiser & Goldammer, 2010: BAMS, 91, SC2009• Benedetti et al., 2011: BAMS, 92, SC2010• Huneeus et al., 2011: ACPD, 11, 7781-7816

• TM653: Simulations of volcanic plumes with the ECMWF/MACC aerosol system• TM659: Prognostic aerosols: MACC vs. GEMS• TM660: Aerosol-Cloud-Radiation Interactions in the ECMWF/MACC forecasts

Submitted* Benedetti et al., 2012: BAMS, 93, State of the Climate 2011* Morcrette et al., 2012: JGR, direct/indirect aerosol effects

http://www.gmes-atmosphere.eu the MACC/MACC II web site